不同神经网络模型评估南水北调中线高填方渠道边坡稳定性

Stability Evaluation of a High Fill Channel Slope in the South-to-North Water Diversion Middle Route Project Based on Different Neural Network Models

在水利工程中,边坡安全稳定性一直是工程中所关注的热点.保障渠道边坡的稳定性对南水北调中线工程安全运行具有重要意义.神经网络方法可以很好地评估边坡稳定性,不同的神经网络模型的精度和计算效率不同,对比BP神经网络模型、GA-BP神经网络模型以及MIV-GA-BP神经网络模型,并根据南水北调中线某高填方渠道边坡的相关数据资料,以有限元正分析所得数据作为样本进行计算,对高填方渠道边坡稳定性进行评估,并取南水北调中线工程实例进行验证.研究表明,BP模型计算误差较大,GA-BP和MIV-GA-BP模型均可较好预测边坡稳定性,其中MIV-GA-BP具有更高的精度与计算效率,且能获取输入参数的敏感性.

In water conservancy engineering,the safety of channel slope has always been the focus of attention in engineering.Ensuring the stability of the channel slope is of great significance to the safe operation of the South-to-North Water Diversion Middle Route Project.The neural network method can evaluate the slope stability well,and the accuracy and calculation efficiency of different neural network models are different.In this paper,BP neural network model,GA-BP neural network model and MIV-GA-BP neural network model are applied and compared.Based on the relevant data of a high fill channel slope in the South-to-North Water Diversion Middle Route Project,the data calculated by finite element software are taken as samples for calculation,the stability of the high filling channel slope is evaluated.Then an example of the South-to-North Water Diversion Middle Route Project is taken for verification.Research shows that the BP model calculation error is bigger,while the GA-BP and MIV-GA-BP model can better predict the slope stability.The MIV-GA-BP has higher accuracy and computational efficiency,and can obtain the sensitivity of the input parameters.